뉴스로 돌아가기
산업AI Understanding 브리핑

Baseten, 통합 에이전트 AI 인프라 플랫폼 구축을 위해 Blaxel 인수

Baseten은 모델 추론 및 훈련 인프라를 자율 AI 에이전트를 위한 실행, 저장 및 네트워킹과 결합하기 위해 Blaxel을 인수했습니다. 재정 조건은 공개되지 않았으며 결합된 플랫폼에 대한 제품 액세스 및 가격은 아직 알려지지 않았습니다.

4 min readRead the linked source
Source-page capture accompanying Baseten acquires Blaxel to build integrated agentic AI infrastructure platform
소스 참조녹음된 소스
출판사
pulse2.com
소스 링크
pulse2.comhttps://pulse2.com/baseten-acquires-blaxel-to-build-integrated-agentic-ai-infrastructure-platform/
소스 유형
연결된 소스 — 기본 소스 상태가 설정되지 않았습니다.
맥락60초 안에 이해하세요

여기서 시작하세요

주요 용어

MCP(모델 컨텍스트 프로토콜)
AI 애플리케이션이 표준 방식으로 외부 도구, 데이터 소스 및 컨텍스트 제공자에 연결할 수 있게 해주는 개방형 프로토콜입니다.
훈련 후
명령어 튜닝, 선호도 최적화, 안전 튜닝 등 사전 학습 이후 적용되는 학습 단계입니다.
추론
훈련된 모델이 예측 또는 출력을 생성하는 런타임 단계입니다.
자신을 테스트해 보세요AI 에이전트 퀴즈

무슨 일이 일어났나요?

Pulse 2.0 reports that Baseten acquired Blaxel, bringing together Baseten’s model training and infrastructure with Blaxel’s stateful execution, storage and networking technology for autonomous agents. The companies plan to develop one platform for training and serving models while running long-lived agents. Existing Blaxel products and support are expected to continue for now, with Sandboxes identified as the first capability Baseten plans to expand.

Pulse 2.0 reports that Baseten acquired Blaxel, combining Baseten’s infrastructure for training and serving AI models with Blaxel’s infrastructure for running autonomous agents. The financial terms were not disclosed. The stated goal is an integrated system in which developers can train and serve models while operating long-running agents in persistent environments.

According to Pulse 2.0, Blaxel’s technology includes Sandboxes: isolated micro-virtual-machine environments where agents can write and execute code; Agent Drive, a distributed filesystem for preserving files, code and working context; and networking for controlled communication with tools, APIs, Model Context Protocol servers and other agents. The outlet reports that Blaxel says its sandboxes can suspend and resume in about 25 milliseconds and remain idle at close to zero computing cost, but these claims were not independently tested in the supplied material.

Pulse 2.0 reports that existing Blaxel customers will see no immediate changes to current products and support, and that the existing team will remain in place. Baseten plans to introduce additional products using Blaxel’s infrastructure primitives, beginning with Sandboxes. The report also says Sapiom runs hundreds of millions of agent loops on Blaxel infrastructure, citing the announcement.

소스 세부정보: pulse2.com ↗

왜 중요한가요?

The deal targets a concrete infrastructure gap in agentic AI: agents need more than repeated model calls. They may execute code, call tools and APIs, preserve files and state, and operate over extended periods. Bringing closer to agent execution, storage and networking could reduce operational complexity and network dependence for developers. The report does not independently confirm the companies’ performance claims, customer scale, or whether the planned integrated platform is generally available.

Agentic applications create infrastructure requirements that differ from conventional . An agent may repeatedly call a model, execute code, interact with external systems and retain state across sessions. A platform that colocates those functions could simplify deployment and potentially improve control over latency, security and costs, although the report provides no independent measurements demonstrating those benefits.

The acquisition also extends Baseten’s stated role from model infrastructure toward the surrounding execution layer. Pulse 2.0 reports that the companies envision connecting agent activity and outputs to workflows, but it is not clear how this would work in production, what safeguards would apply, or whether customers will receive such capabilities.

Interactive Mechanism

대화형 메커니즘: 실제로 작동하는 방식

이 개발의 이면에 있는 기본 기술을 대화식으로 살펴보세요.

Agent Lifecycle Stage:
1
User Intent & Planning: "Audit customer refund request #4092 and settle payment."
2
Tool Calling: Emits structured JSON call crm_get_transaction(id='4092').
3
Guardrail & Verification:🛡️ Paused: High-value action requires human operator sign-off.
4
Final Settlement: Refund recorded, email receipt dispatched, and audit log stored.
Core takeaway: An AI agent is not just a language model—it is a closed loop of planning, tool invocation, and environment feedback. Production systems require self-healing retries and strict human approval guardrails.
대화형 개념 확인+10 Points
AI Agents Quiz

An agent must create a draft calendar event for Tuesday at 2 p.m. Which evidence would establish the requested result?

다음에 무엇을 볼 것인가

The main uncertainties are execution, availability and economics. Watch for Baseten’s product announcements about Sandboxes and the combined platform, documentation showing who can access them, pricing and regional coverage, and independent evidence about latency, isolation, reliability and cost. It is also unclear whether the acquisition will produce a broadly available service or remain limited to existing customers and selected deployments.

No public access terms, pricing, launch timetable or general-availability status are provided in the supplied report. The companies’ longer-term objective of supporting millions of autonomous agents remains a plan, not a demonstrated result. Future evidence should clarify which features are available, to whom, in which regions, and under what isolation, data-retention and security controls.

Independent testing will also be important for the reported sandbox speed, idle-cost behavior, reliability and workload isolation. The supplied source does not establish how the combined platform compares with alternatives or whether customers can migrate existing agent workloads without changes.

관련 가이드 및 퀴즈

AI 에이전트AI 모델 설명AI 트레이닝AI 윤리알고 있는 내용을 테스트해 보세요. 무료 AI 퀴즈를 시도해 보세요.용어집에서 AI 용어를 찾아보세요.AI 자금 추적기를 팔로우하세요
이것이 유용하다고 생각하시나요?